Tidypredict with recipes, turn workflow to SQL, spark, duckdb and beyond
summary
A new package called orbital package turns a fitted tidymodels workflow (including recipes) into a compact object that generates equivalent predictions, SQL or JavaScript code and a standalone R function; predictions match the original model to many decimal places and it exports directly to SQL, Spark, DuckDB. It implements about 46 steps of recipes, produces tiny deployable artifacts, and can run predictions inside databases or in-browser, but some model types (e.g. nearest neighbors) and input validation are not supported.
AI-written summary · about
chapters (11)
- 0:00Tidypredict and Orbital intro
- 2:05Tidypredict workflow limits
- 4:15Tree-based model predictions
- 8:01Orbital for efficient predictions
- 9:30Predictions in databases
- 10:55SQL from Orbital objects
- 12:03Shiny app deployments
- 13:53Predictions to JavaScript
- 15:01Orbital unsupported models
- 16:47Orbital advantages and future
- 17:12Q&A
related talks
57mGood Practices for Applied Machine Learning- 57mGood Practices for Applied Machine Learning
18mtidymodels: Adventures in Rewriting a Modeling Pipeline
15mThe tidysynthesis R package
25mSql Server and R for Real-time Predictions
31mCreating recommender systems with Spark, Scala and Prediction.Io
19mtidyclust - expanding tidymodels to clustering
23mparsnip A tidy model interface